Vehicle Re-Identification in Surveillance Environment

Abstract

An intelligent transportation system (ITS) usually needs to automatically monitor the road traffic, and may also need to detect and track vehicles of interest. This paper is concerned with an extension of ITS to a campus surveillance environment to identify vehicles in images in different locations of a campus. The proposed approach uses corners of side windows as image features of a vehicle and maps them to a nominal configuration under which relative positions among them are compared. Experimental results show that the similarity measures thus established can indeed achieve the objective of vehicle identification effectively, and in an efficient way.